Kyle Gwinnup

program analysis, compilers, bayesian, founder at

[Kyle Gwinnup @geom_kpg]

I love Kaplan-Meier curves. They shouldn't require code, and I shouldn't have to leave my spreadsheet to make one.

So in TukeySheets: select three columns, insert survival chart. That's it. #stats #dataviz #science

[Kyle Gwinnup @geom_kpg]

@RyPeck It’s a secret for a couple more weeks

[Kyle Gwinnup @geom_kpg]

Moved both companies to self-hosted #Forgejo + runners behind #Tailscale. We gained four things: platform stability, data privacy, full runner control (better specs, number of runners), and Tailscales VPN that just works. A joy to dev with. Can't believe I never did this earlier.

[Kyle Gwinnup @geom_kpg]

@twtayaan My companies started self hosting #forgejo and #forgejo runners . AI makes it easier than ever to do this. Highly recommend it (self hosting and #forgejo) You also now own your data.

[Kyle Gwinnup @geom_kpg]

Programmers love to dunk on spreadsheets. Meanwhile spreadsheets run more of the world's analysis than every notebook and BI dashboard combined.
The form isn't the problem. https://t.co/LjghUlgAV1 #science #stats #dataviz

[Kyle Gwinnup @geom_kpg]

https://t.co/8PtaqSmWuw Survival analysis now lives in the spreadsheet, no exporting to R or SPSS. Kaplan-Meier curves, log-rank tests, and hazard ratios are just SURV.* formulas. The data, the formulas, the chart stay in one document, edit a cell, the curve updates. #statistics

[Kyle Gwinnup @geom_kpg]

Had some great feedback, and made the chart picker a little clearer in the format bar. There is now an icon, text, and chevron to select a chart. They also stagger when you add multiple so one never hides the other. Thank you, and keep it coming :) #Stats #Data

[Kyle Gwinnup @geom_kpg]

big progress in pancreatic cancer treatment! https://t.co/xgVPV8rTDr
#cancer #research

[Kyle Gwinnup @geom_kpg]

Launching TukeySheets, a new age spreadsheet.

New classes of stats formulas, typst integration and editor, flexible layered plotting, database querying, AI assistant that understands your sheet GPU rendered native app! https://t.co/LjghUlgAV1 feedback is welcome #stats #dataviz

[Kyle Gwinnup @geom_kpg]

Two things I believe: (1) the spreadsheet is the best interface humans have for data. (2) the one everyone uses is holding us back. Both are true. So I'm fixing #2. → https://t.co/LjghUlgAV1 #stats #dataviz

[Kyle Gwinnup @geom_kpg]

After years bouncing between R/Python and spreadsheets, I prefer the spreadsheet, and started building the R/Python parts into it.

https://t.co/Gt1wlx5SZF ggplot-style layered plots, direct warehouse queries. AI-native. In closed beta now, open beta soon!
#RStats

[Yaron (Ron) Minsky @yminsky]

A gem from Stephen Dolan, which proposes replacing the "generational hypothesis" that drives the design of generational GCs with a notion of lifetime dispersion as measured by the gini coefficient. Nice to see economics playing a role here!

https://t.co/BbM18XsxG0

[Richard McElreath 🐈‍⬛ @rlmcelreath]

Kinds of science

True & Replicable: photoelectric effect

False & non-Replicable: many priming results

F & Replicable: bad analyses like confounded country-level regressions

T & non-Replicable: badly documented experiments, unique observations

I think there's a lot of F&R

[Stuart Buck @stuartbuck1]

The constant reinventing-the-wheel in AI/ML (and the needless relabeling of old concepts from statistics) seems like an impediment to making progress and communicating with other people.

When a statistician as brilliant as Cosma Shalizi says this . . .

https://t.co/qCKHosBu0B

[Zhengzhong Tu @_vztu]

Reinforcement learning explained

https://t.co/hs4pE5fS2r